Gartner expects the market to grow 63.4% this year, with the fastest growth coming from smaller, specialized models rather than headline-grabbing Frontier systems.
Global end-user spending on AI models and platforms will total USD 64.25 billion in 2026, an increase of 63.4% from USD 39.31 billion last year, according to a forecast released by Gartner on July 20. This growth rate is high no matter how you look at it. The details below describe a market where buyers are starting to ask tougher questions about what they’re getting in return.
“Enterprise AI budgets are coming under increased scrutiny, with an increased focus on utilization efficiency, cost control, and measurable outcomes,” said Arunasree Cheparthi, senior principal research analyst at Gartner. He said spending is going toward providers that can demonstrate value in terms of cost, latency, performance and reliability.
Gartner divides the market into two. This year, spending on the models themselves will increase by 117%, while spending on the platforms used to build and run them will increase by 36.9%. Models will account for about 44% of the total in 2026, up from a third in 2025.
Among the model halves, the underlying generative AI systems still account for the majority of spending, increasing 104.2% to USD 23.36 billion from USD 11.44 billion in 2025. The platform halves are split between AI application development platforms, which are expected to grow by 38.6% to reach US$9.54 billion, and platforms for data science and machine learning, which are expected to grow by 36.3% to reach US$26.44 billion.
That last number makes data science and machine learning platforms the largest single segment in the market, surpassing underlying models. This category covers software that organizations use to prepare data, train models, and put them into production, and predates the current generative AI cycle by several years. Although its growth rate is the slowest of the four segments, it is still the layer on which the rest of the market is built.
The fastest growing segment is also the smallest
Domain-specific language models (DSLM) are expected to grow 210% this year, increasing from US$1.58 billion to US$4.91 billion. This percentage is significant primarily due to the low starting point. Even if it tripled, this sector would still account for less than 8% of total spending.
DSLMs are trained or tailored for a narrow purpose rather than a general competency, covering tasks such as reading clinical records, processing insurance claims, and answering questions about specific product catalogs. They are smaller than frontier models, so they cost less to run and are faster to respond to. In many cases, it can be hosted on the hardware an organization already has in operation, rather than being accessed through a third-party interface.
This is important in a market where data retention rules limit what can be sent outside the organization and where AI budgets are measured more closely than even the largest cloud providers in the US or China. A model that has low cost per query and stays behind the firewall answers two constraints simultaneously. This is a big part of why Gartner expects the segment to triple.
Small slice on a much larger bill
The US$64 billion covers models and the software used to build and manage them. It does not cover the underlying hardware. Gartner separately predicts that total global AI spending will exceed USD 2.5 trillion in 2026, with spending on data center systems alone expected to exceed USD 788 billion this year.
In this comparison, the model and platform layers will account for approximately 2.5% of the money organizations will spend on AI in 2026. Most of the money is still spent on buildings, power, and chips.
Ceparti’s argument is that the commercial advantage is shifting to suppliers who make the use of AI more transparent to those paying them. She pointed to providers that incorporate ratings, cost transparency, and usage tracking into their customers’ workflows. This field has grown into a category of its own as organizations seek to run multiple models in parallel and route each task to the cheapest model that can handle it.
The same shift creates problems for sellers. When pricing moves from fixed licenses to consumption, revenue depends on continued usage rather than renewals. “As more models come to market and usage-based pricing becomes less predictable, buyers will look to platforms that help them choose the right tools, monitor performance, enforce policy, and control costs,” Ceparti said.
Gartner’s own research warns about how much of projected spending turns into actual working systems. The company predicts that more than 40% of agent AI projects will be canceled by the end of 2027 due to rising costs, unclear business value, and inadequate risk management. Expenditure forecasts count what buyers commit to, not what survives contact with production.
